Roles of Residues Arg-61 and Gln-38 of Human DNA Polymerase η in Bypass of Deoxyguanosine and 7,8-Dihydro-8-oxo-2′-deoxyguanosine*

Background: Arg-61 and Gln-38 of human DNA polymerase (hpol) η play important roles in the catalytic reaction. Results: Mutations R61M or Q38A/R61A dramatically disrupt the activity of hpol η. Conclusion: Polarized water molecules can mimic and partially compensate for the missing side chains of Arg-61 and Gln-38 in the Q38A/R61A mutant. Significance: The positioning and positive charge of Arg-61 synergistically contribute to the activity of hpol η, with additional effects of Gln-38. Like the other Y-family DNA polymerases, human DNA polymerase η (hpol η) has relatively low fidelity and is able to tolerate damage during DNA synthesis, including 7,8-dihydro-8-oxo-2′-deoxyguanosine (8-oxoG), one of the most abundant DNA lesions in the genome. Crystal structures show that Arg-61 and Gln-38 are located near the active site and may play important roles in the fidelity and efficiency of hpol η. Site-directed mutagenesis was used to replace these side chains either alone or together, and the wild type or mutant proteins were purified and tested by replicating DNA past deoxyguanosine (G) or 8-oxoG. The catalytic activity of hpol η was dramatically disrupted by the R61M and Q38A/R61A mutations, as opposed to the R61A and Q38A single mutants. Crystal structures of hpol η mutant ternary complexes reveal that polarized water molecules can mimic and partially compensate for the missing side chains of Arg-61 and Gln-38 in the Q38A/R61A mutant. The combined data indicate that the positioning and positive charge of Arg-61 synergistically contribute to the nucleotidyl transfer reaction, with additional influence exerted by Gln-38. In addition, gel filtration chromatography separated multimeric and monomeric forms of wild type and mutant hpol η, indicating the possibility that hpol η forms multimers in vivo.

Y-family DNA polymerases are characterized by their low fidelity and ability to tolerate a wide range of lesions during DNA synthesis. Due to their relatively large active sites and loss of 3Ј-5Ј exonuclease domains compared with replicative DNA polymerases, they readily introduce mutations into the ge-nome. These enzymes are able to insert dNTPs opposite and beyond DNA lesions, therefore rescuing blocked replication forks and releasing the arrested cell cycle (1)(2)(3)(4)(5)(6).
Human DNA polymerase (hpol ) 2 belongs to the DNA polymerase Y family (5). It plays a crucial role in bypassing cyclobutane pyrimidine dimers, which arise as a result of UV exposure. Defects in hpol bypass of cyclobutane pyrimidine dimers in xeroderma pigmentosum variant patients results in increased rates of skin and internal cancers (7)(8)(9). hpol contributes to replication past DNA lesions induced by platinum chemotherapeutic drugs, leading to escape of the arrested cell cycle and therefore proliferation of cancer cells (10 -17). Besides translesion synthesis, there is evidence that hpol is involved in homologous recombination by interaction with Rad51 and extension of the D-loop (18 -20). hpol also contributes to somatic hypermutation in B cells by introducing mutations in specific DNA stretches embedded within antigenbinding sites, thereby causing structural changes and potentially elevating the antigen-binding ability (21)(22)(23)(24)(25)(26)(27).
hpol and its yeast homolog Rad30 also play roles in the bypass of 7,8-dihydro-8-oxoguanine (8-oxoG) (28 -30), one of the most common DNA lesions introduced by oxidative stress. 8-oxoG is generated at a rate of Ͼ1000 molecules/cell/day and prefers the syn conformation, which facilitates formation of a Hoogsteen pair with dATP and introduces mutations into the genome (31)(32)(33). hpol replicates past 8-oxoG mainly by inserting dCTP, scaffolding the dCTP:8-oxoG pair in the Watson-Crick geometry, although the dATP misincorporation rate is 280-fold higher in the bypass of 8-oxoG compared with unmodified G (30).
Crystal structures of the catalytic core of hpol (amino acids 1-432) and its yeast homolog with unmodified or lesion-containing oligonucleotides have been published (10, 11, 21, 28, 30, 34 -39). These studies provide a better understanding of the catalytic mechanism and have yielded valuable information regarding the design of potential drugs against hpol in the hope of blocking tumor proliferation. The residue Arg-61, located in a short loop region, is highly conserved. Its dynamic movement contributes to the catalytic activity of the enzyme. X-ray crystal structures reveal that after binding the DNA substrate in the absence of an incoming dNTP, the positively charged guanidino moiety of Arg-61 is projected toward the nucleobase of the 5Ј-overhanging, single-stranded region of template DNA of the template-primer duplex (38). However, following binding of dNTP and divalent metal ions, the side chain of Arg-61 undergoes a rotation and interacts with the ␣-phosphate of the incoming dNTP. Interestingly, the side chain of Arg-61 also moves during phosphodiester bond formation to allow enough space for a third divalent metal ion, which interacts with both the ␣-phosphate and the oxygen, bridging the ␣and ␤-phosphates. Thus, Arg-61 can adopt different conformations during each stage of the catalytic reaction cycle. In addition, the reported structures have also indicated that another highly conserved residue near the active site, Gln-38, may play an important role in stabilizing the template base in the nucleotidyl transfer reaction (10,11,21,30,34,35,37).
Site-directed mutagenesis was used in this study to further investigate the roles of Arg-61 and Gln-38 in the catalytic activity. Wild-type and mutant forms of hpol were analyzed and crystallized with unmodified DNA or DNA containing 8-oxoG. The combined results of the structural and biochemical analysis indicate that the positioning and positive charge of the Arg-61 synergistically contribute to the nucleotidyl transfer reaction in addition to the hydrogen bonding effect of Gln-38 with the template base. When both Arg-61 and Gln-38 were replaced by Ala, polarized water molecules could mimic and partially compensate for the missing wildtype side chains.

Experimental Procedures
Materials-All oligonucleotides were purchased from Integrated DNA Technologies (Coralville, IA), and some were purified by HPLC. dNTPs were from New England Biolabs (Ipswich, MA). PreScission protease was from GE Healthcare. Polyethylene glycol monomethyl ether 2000 (for crystallization) was from Hampton Research (Aliso Viejo, CA).
Solution B contained 1 mM dCTP and 10 mM MgCl 2 . The two solutions were mixed rapidly with a KinTek RP-3 instrument (KinTek Corp., Austin, TX) at 25°C for 0.005-5 s (the final reaction concentration for wild-type hpol was 50 nM, and that for mutant hpol was 250 nM) and stopped by the addition of 0.5 M EDTA. The products were separated with 18% denaturing polyacrylamide gels (w/v) and visualized with a Typhoon system (GE Healthcare). After quantification using ImageJ software, the results were fit to a burst equation using GraphPad Prism: y ϭ A(1e Ϫkpt ) ϩ k ss E 0 t, where A is the burst amplitude, representing the apparent concentration of the active form of the enzyme, k p is the burst rate, k ss is the steady-state rate, and E 0 is the total enzyme concentration (30,37,40).
Crystallizations-The hanging drop vapor diffusion technique was used to obtain crystals. DNA complex and protein (1.2:1 molar ratio) were first incubated at room temperature for 10 min, followed by the addition of 0.5 l of 1 M CaCl 2 and concentration by centrifugation through a filter (10 kDa, Millipore). After the addition of 0.5 l of 100 mM dNTP (New England Biolabs), the solution (1 l) was mixed with an equal volume of reservoir solution (100 mM sodium MES (pH 6.0), 5 mM CaCl 2 , and 15-22% (w/v) polyethylene glycol monomethyl ether 2000 (Hampton Research, Aliso Viejo, CA)). The mixture was equilibrated against 500 l of the reservoir solution at 18°C. Crystals were typically observed after overnight incubation (30,35).
X-ray Diffraction Data Collection and Structure Determination-Crystals were mounted in nylon loops and swiped through a mixture of reservoir solution plus 25% glycerol (v/v), followed by flash-cooling in liquid nitrogen. Diffraction data were collected on the Sector 21-ID-G beamline at the Advanced Photon Source (Life Sciences Collaborative Access Team, Argonne National Laboratory, Argonne, IL). All collected data were integrated and scaled using HKL2000 (41). Structures were determined by molecular replacement phasing using the program Phaser MR (42). Coordinates with Protein Data Bank (PDB) code 4O3N were used as the search model for R61M⅐G⅐dCTP and Q38A/R61A⅐G⅐dCTP, those with PDB code 4O3P were used for R61M⅐8-oxoG⅐dCTP and Q38A/R61A⅐8-oxoG⅐dCTP, and those with PDB code 4O3O were used for R61M⅐8-oxoG⅐dATP. PHENIX (43) was applied for refinements, and ARP/ ARP classic (44,45) and COOT (46) were used for model building. Structural illustrations were generated with the program UCSF Chimera (47).

Primer Extension past G or 8-oxoG by Wild-type or Mutant hpol in the Presence of All Four dNTPs-
The catalytic core (residues 1-432) of hpol was used in this study because its activity is similar to that of the full-length protein in vitro (10). In order to understand the mechanism of dNTP incorporation by hpol , point mutations were introduced at Arg-61 and Gln-38. In the extension assays, wild-type hpol bypassed both unmodified template G and 8-oxoG and fully extended the primer in 15 min, whereas mutants R61M and Q38A/R61A partially elongated the primer by inserting at most two nucleotides, indicating that these mutations strongly attenuate the activity of hpol . In comparison, the mutations Q38A, Q38L, R61A, and R61K only slightly reduced the catalytic activity ( Fig.  1, A and B).
Steady-state Kinetics of dCTP or dATP Incorporation Opposite G or 8-oxoG by Wild-type and Mutant hpol -All hpol mutants used in our study had attenuated efficiency for dCTP insertion opposite unmodified G compared with the wild-type enzyme. The Q38L, R61M, and Q38A/R61A mutants showed 22-, 19-, and 16-fold attenuation in enzyme efficiency, respectively, whereas Q38A, R61A, and R61K showed only small decreases. Even when these mutations were introduced, hpol still maintained relatively high fidelity opposite G, and the dATP misincorporation frequency for each hpol was Ͻ4% ( Fig. 2A and Table 1). For dCTP incorporation opposite 8-oxoG ( Fig. 2B and Table 2), the lowest catalytic efficiencies (k cat /K m ) were observed for the Q38L, R61M, and Q38A/R61A mutants

Arg-61 and Gln-38 Roles in Human Polymerase
(i.e. 4.9, 1.4, and 2.2 M Ϫ1 min Ϫ1 respectively, compared with 31 M Ϫ1 min Ϫ1 for wild-type hpol ). In the study of steadystate kinetics of dATP misinsertion opposite 8-oxoG, a second insertion band was always observed due to the next base T, 3Ј to 8-oxoG on the template strand (Fig. 2C). To obtain catalytic efficiencies specifically for the insertion step opposite 8-oxoG, the two insertion bands were both included in the quantitation. We reported previously that the dATP insertion frequency was 280-fold higher for wild-type hpol replicating past 8-oxoG than that for replicating past G (30). For each mutant, dATP was more frequently inserted opposite 8-oxoG than G. Interestingly, all mutants except R61K showed higher fidelity (inserting dCTP instead of dATP) than wild-type hpol in bypassing 8-oxoG. The R61K mutant, consistent with a previous report (21), had a higher misincorporation frequency (0.60, compared with 0.28 for wild-type hpol ) ( Fig. 2B and Table 2). Pre-steady-state Kinetics of dCTP Incorporation Opposite G or 8-oxoG by Wild-type and Mutant hpol -In the pre-steadystate kinetic analysis of dCTP insertion opposite G, the burst rates for R61M and Q38A/R61A were 2.6-and 5.9-fold lower than that for wild-type hpol ( Fig. 3 and Table 3 Enzyme efficiencies for dNTP incorporation opposite G or 8-oxoG by wild-type or mutant hpol . Enzyme efficiencies (k cat /K m ) were obtained from steady-state kinetic studies by incubating 2-500 nM hpol , 5 M FAM-labeled primer⅐template DNA complex, and varying concentrations of dNTPs at 37°C for 5 min. After separating the products on denaturing polyacrylamide gels, data were fit to a (hyperbolic) Michaelis-Menten equation using non-linear regression (Prism). Each experiment was conducted at least twice, and error bars indicate S.D. A, dCTP and dATP incorporation opposite G; B, dCTP and dATP incorporation opposite 8-oxoG; C, an example (R61A mutant) of second band insertion during a steady-state-kinetic study for dATP incorporation opposite 8-oxoG.

TABLE 1 Steady-state kinetics of incorporation of dCTP and dATP opposite G by wild-type and mutant hpol
The oligonucleotides used were 5Ј-FAM-CGGGCTCGTAAGCGTCAT-3Ј and 3Ј-GCCCGAGCATTCGCAGTAGTACT-5Ј.
For the bypass of 8-oxoG, the burst amplitudes were ϳ14% for each hpol mutant and ϳ70% for wild-type hpol . Notably, for each individual enzyme, the burst amplitude for incorporation opposite G was always larger than that for 8-oxoG, suggesting that non-productive forms are more prominent in the presence of 8-oxoG (48) ( Table 3).

Post-G or -8-oxoG Synthesis by Wild-type or Mutant hpol -
The ternary crystal structure shows that 8-oxoG formed a Watson-Crick pair with an incoming dCTP analog in the presence of wild-type hpol (PDB code 4O3P) (30). The question can be raised as to whether the presence of an 8-oxoG:C pair affects primer extension by wild-type or mutant hpol . In the presence of all four dNTPs, wild-type hpol fully extended the primer beyond the 8-oxoG:C pair in 15 min, and the Q38A, Q38L, R61A, and R61K mutants elongated the primer at slightly slower rates, but only part of the primer was extended (by one nucleotide) by the R61M and Q38A/R61A mutants (Fig. 4A).
To further investigate primer extension post-8-oxoG, steady-state kinetic analysis was applied for the next base (dATP) insertion beyond an 8-oxoG:C pair, as well as a G:C pair as a control. Consistent with the observation in Fig. 4A, the catalytic efficiencies (k cat /K m ) for dATP insertion post-8-oxoG were much lower for R61M (1.8 M Ϫ1 min Ϫ1 ) or Q38A/R61A (1.2 M Ϫ1 min Ϫ1 ) compared with wild-type hpol (47 M Ϫ1 min Ϫ1 ). Interestingly, both wild-type hpol and the two mutants had similar k cat /K m values for dATP incorporation post-G and -8-oxoG, indicating that 8-oxoG does not affect next base insertion by hpol , even with mutations at Arg-61 or Gln-38 (Table 4).
Because wild-type or mutant hpol may incorporate dATP opposite 8-oxoG, extension assays beyond the 8-oxoG:A pair were also studied. The extension pattern beyond an 8-oxoG:A pair was very similar to that beyond an 8-oxoG:C pair for wildtype hpol or either mutant, indicating that the mismatched 8-oxoG:A pair does not dramatically affect subsequent dNTP insertion (Fig. 4B).
X-ray Crystallography of Mutant hpol Inserting dCTP Opposite G-In order to further understand the roles of Arg-61 and Gln-38 in bypassing unmodified G, crystal structures of R61M⅐G⅐dCTP (R61M mutant incorporating dCTP opposite unmodified template G) and Q38A/R61A⅐G⅐dCTP were obtained in the presence of Ca 2ϩ , which does not support nucleotidyl transfer reactions (Table 5). Electron density for the active site of each ternary complex is shown in Fig. 5. Similar to the structure of the wild-type hpol ⅐G⅐dCTP complex, hpol mutants R61M and Q38A/R61A provided a scaffold for the incoming dCTP and template G to form a Watson-Crick base pair (Fig. 6). The distances were 3.8 and 4.1 Å from the 3Ј-OH of T of the primer strand to P ␣ of the incoming dCTP in the

TABLE 3 Burst kinetics of incorporation of dCTP opposite G and 8-oxoG
The oligonucleotides used were 5Ј-FAM-CGGGCTCGTAAGCGTCAT-3Ј and 3Ј-GCCCGAGCATTCGCAGTAGTACT-5Ј; 5Ј-FAM-CGGGCTCGTAAGCG-TCAT-3Ј and 3Ј-GCCCGAGCATTCGCAGTA(8-oxoG)TACT-5Ј. R61M⅐G⅐dCTP and Q38A/R61A⅐G⅐dCTP complexes respectively, compared with 3.3 or 3.5 Å (two conformations) for wildtype hpol (30). Arg-61 of wild-type hpol engages in favorable Coulombic interaction with the ␣-phosphate of the incoming dCTP. The electrostatic force was lost after mutating it to the hydrophobic Met, although the size of the side chain was similar to Arg. Interestingly, in the ternary Q38A/R61A⅐G⅐dCTP complex, one water molecule was located above the base ring of dCTP and likely to form a lone pairor H-interaction with the nucleobase (Fig. 6) (49).
In the minor groove, the N ⑀ atom of Gln-38 connected with N3 of the template G through a hydrogen bond and a water molecule-bridged O ⑀ of Gln-38 and N2 of G, thus stabilizing the template G to form a base pair with the incoming dCTP. When Gln-38 was mutated to Ala, these interactions were disrupted, and the bridging water molecule was not present in the Q38A/ R61A⅐G⅐dCTP structure (Fig. 6).
X-ray Crystallography of Mutant hpol Inserting dCTP Opposite 8-oxoG-To investigate the roles of Arg-61 and Gln-38 in bypassing 8-oxoG, mutant hpol was crystallized with dCTP opposite 8-oxoG in the presence of Ca 2ϩ . Fourier (2F o Ϫ F c ) sum electron density maps are shown in Fig. 5. Similar to the structures depicting dCTP insertion opposite unmodified G, wild-type hpol and the mutants provided a scaffold for dCTP pairing with 8-oxoG in the Watson-Crick geometry, with 8-oxoG in the anti conformation (Fig. 7). The distances between the primer terminal 3Ј-OH and the P ␣ of dCTP were 3.2 or 3.4 Å in the complex with wild-type hpol (30), 3.3 or 3.9 Å with the R61M mutant, and 3.9 Å with Q38A/ R61A, respectively. Similar to dCTP insertion opposite G by wild-type hpol , the side chain of Arg-61 is involved in an electrostatic interaction with the ␣-phosphate of dCTP in the wild-type hpol ⅐8-oxoG⅐dCTP complex, but this interaction was destroyed when replacing Arg-61 with Met. Interestingly, when Arg-61 was mutated to Ala, water molecules were lodged on top of dCTP. One of the water molecules was on top of the nucleobase of the dCTP, whereas the position of the other matched that of N of Arg-61 from the wild-type hpol ⅐8-oxoG⅐dCTP complex, allowing it to donate a hydrogen bond to the ␣-phosphate of the incoming dCTP (Fig. 7).
In the minor groove of the wild-type ternary complex, N ⑀ of Gln-38 was engaged in a hydrogen bond with N3 of 8-oxoG. After superimposition, a water molecule in the Q38A/R61A⅐8-oxoG⅐dCTP complex was observed near the N ⑀ of Gln-38 in the wild-type hpol ⅐8-oxoG⅐dCTP complex. Further investigation revealed a hydrogen bonding interaction between this water and the N3 of 8-oxoG. In addition, in the wild-type hpol ⅐8-oxoG⅐dCTP complex, a water molecule bridged both the O ⑀ of Gln-38 and N2 of 8-oxoG by hydrogen bonds. This bridging water hydrogen-bonds with the N ⑀ -mimicking water as well as N2 of 8-oxoG in the Q38A/R61A⅐8-oxoG⅐dCTP complex (Fig. 7).
X-ray Crystallography of Mutant hpol Inserting dATP Opposite 8-oxoG-To explore the role of Arg-61 in the fidelity of hpol , we crystallized the hpol R61M mutant with dATP opposite 8-oxoG. Similar to the wild-type hpol ⅐8-oxoG⅐dATP complex, 8-oxoG adopted the syn conformation and formed a Hoogsteen pair with the incoming dATP when Arg-61 was mutated to Met. The distance from the 3Ј-OH of the primer end to P ␣ of dATP was 3.7 Å, com-

Arg-61 and Gln-38 Roles in Human Polymerase
The methionine side chain in the point mutant R61M exhibited two conformations in the ternary structure. One of the conformations partially superimposed with the side chain of Arg-61 in the wild-type hpol ⅐8-oxoG⅐dATP complex. However, due to the hydrophobic nature of the Met side chain, the electrostatic interaction seen in the wild-type complex with the incoming dATP was lost. The N ⑀ of Gln-38 from either wildtype or R61M mutant hpol donated a hydrogen bond to the O8 atom of 8-oxoG in the minor groove (Fig. 8).
Multimeric Forms of hpol -Wild-type hpol and the mutants R61M and Q38A/R61A were purified by gel filtration chromatography (Superdex-75 10/300 GL) prior to crystallization. Two distinct peaks, eluting with t R of ϳ8.1 and 11.3 ml, were observed in the chromatogram in each case, suggesting the presence of proteins with different molecular masses (Fig.  9A). The second peak was eluted with the molecular mass of monomeric hpol , as established using the standards BSA (67 kDa) and ovalbumin (43 kDa). The proteins from the two peaks had the same mobility in SDS-PAGE (data not shown), indicating that the protein in the first peak was a multimeric form of hpol . The R61M mutant had the largest ratio of peak 1 (multimer) over peak 2 (monomer) (Fig. 9A), and kinetic assays with minimal protein dilution following the chromatography (final dilution ratio 1:4 (v/v) for the reaction) were used to measure their catalytic activities (burst phase, pre-steady state). The protein from peak 2 (monomer) showed a 1.5-fold higher burst rate compared with that from peak 1 (multimer) for dCTP insertion opposite template G. The burst amplitudes were 65 and 202 nM (compared with the input concentrations of 566 and 772 nM for the peak 1 and 2 proteins), respectively. In addition, both peaks of the R61M mutant were crystallized with DNA and incoming dCTP with no dramatic structural differences (data not shown). To characterize the multimeric composition, the mutant R61M, as well as two markers, thyroglobulin (670 kDa) and BSA (67 kDa), were used to calibrate a Superdex-200 10/300 GL gel filtration column. The mutant enzyme was eluted in a peak corresponding to monomeric hpol , a broad peak with molecular mass in the range of 400 -600 kDa, and the void peak, indicating the presence of an even higher fold ensemble (Fig. 9B).

Discussion
The fidelity of some DNA polymerases is determined largely by a single key residue. For example, Arg-332 in the archbacterial Sulfolobus solfataricus Y-family DNA polymerase Dpo4 directly interacts with the O8 atom of 8-oxoG and determines its fidelity in the bypass of 8-oxoG (50). Structures of hpol show that two highly conserved residues, Arg-61 and Gln-38, are important in the catalytic reaction (10, 11, 21, 28, 30, 34 -39). In order to delineate their roles for the activity and fidelity of hpol , we mutated the two residues to amino acids with different side chain properties. The unmodified G and a common DNA lesion, 8-oxoG, were used in DNA templates in our study.
The misincorporation frequencies of wild-type and hpol mutants for dATP insertion opposite unmodified G were very small, all Ͻ4%, as judged from the steady-state kinetic assay results. However, the misincorporation frequency was increased 36-fold for mutant Q38L, 13-fold for R61K, 8-fold for Q38A, and 7-fold for Q38A/R61A compared with that for wildtype hpol , whereas mutants R61A and R61M had misincorporation frequencies similar to that of the wild-type protein in incorporating dATP opposite G ( Fig. 2A and Table 1). On the other hand, because 8-oxoG can readily form a syn conformation due to a steric effect and pair with A, the dATP insertion frequency was increased 280-fold for wild-type hpol compared with that opposite unmodified G (30). Interestingly, mutant R61K had the highest misincorporation frequency (0.60) for dATP insertion opposite 8-oxoG, whereas R61M had the lowest value (0.079) compared with wild-type hpol (0.28). The other mutants had slightly reduced misincorporation frequency ( Fig. 2B and Table 2). In addition, previous reports also showed that the template base/lesion as well as the flanking sequence also affected the misincorporation frequency for wild-type or mutant hpol (21,51). Altogether, we conclude that active site residues, the template base/lesion, and the flank-  JUNE 26, 2015 • VOLUME 290 • NUMBER 26

Arg-61 and Gln-38 Roles in Human Polymerase
ing DNA sequence synergistically determine the misincorporation frequency of hpol .
It was unexpected that the R61A mutant would be more active than R61M; however, x-ray crystal structures of the cor-responding complexes provide a potential explanation. Polarized water molecules mimic the side chain of Arg-61 and partially compensate for its functions in the R61A mutant. In order to understand how these water molecules participate in the nucleotidyl transfer reaction, the Q38A/R61A⅐8-oxoG⅐dCTP complex was compared with the other reported wild-type hpol structures. A third metal ion (Mg 2ϩ ) has been reported to be involved in the catalytic process, whereas the side chain of Arg-61 of wild-type hpol adopted the conformation facing toward the major groove ( Fig. 10A) (34). After superimposition, the water interacting with the ␣-phosphate in the Q38A/ R61A⅐8-oxoG⅐dCTP complex partially overlapped with a water molecule coordinating with the third Mg 2ϩ (Fig. 10, B and C). Therefore, the water in the mutant hpol may not only play a role in mimicking the missing side chain of Arg-61 in the ground state but may also interact with the third metal ion during the catalytic process.
Because H 2 O and monovalent Na ϩ may look very similar in electron density maps (52), close investigation of the coordination, occupancy, and B factor was conducted (for those water molecules). In particular, the water molecule on top of the base ring of the incoming dCTP was of interest. This water is 2.9 Å from the centroid of the aromatic ring in the Q38A/ R61A⅐G⅐dCTP complex, and the corresponding distance in the Q38A/R61A⅐8-oxoG⅐dCTP complex is 3.1 Å, indicating the possibility of the formation of a lone pairor H-interaction between H 2 O and the base ring (49). To distinguish H 2 O from a monovalent metal ion, the former was replaced by a Na ϩ ion with half-occupancy in the Q38A/R61A⅐8-oxoG⅐dCTP model   structure. After refinement using the program Phenix, the occupancy of Na ϩ was 0.38, and the B factor dropped from 35.6 to 21 A 2 compared with a fully occupied water (data not presented). Thus, we cannot completely exclude the possibility that the electron density peak represents a partially occupied Na ϩ instead of H 2 O on top of the base ring. Soaking crystals with K ϩ or Rb ϩ and collecting anomalous data may help to further resolve the issue, although K ϩ and Rb ϩ have larger ionic radii than Na ϩ (53).
On the other hand, the electrostatic interaction with the incoming dNTP was lost when Arg-61 was mutated to the hydrophobic Met. In addition, the side chain of R61M was too big to allow room for Arg-61-mimicking water molecules, as seen in the ternary structures with the mutation R61A. As a result, hpol with the single substitution R61A was more active than with R61M. Similar results were observed with the Q38A and Q38L mutants (Tables 1 and 2 and Figs. 6 and 7).
A second unexpected result was observed from the presteady-state kinetic data; the burst amplitudes for wild type hpol inserting dCTP opposite unmodified G or 8-oxG were ϳ100 or 70% of the input, respectively, whereas the amplitude was Ͻ21% for either the R61M or Q38A/R61A mutant. The result indicated the presence of either an inactive form(s) of mutant protein or an equilibrium of reactive and nonreactive ternary complexes (1). Interestingly, gel filtration chromatography showed the presence of both multimeric and monomeric forms of wild type or mutants of hpol . Both forms of the proteins were active and formed single crystals (data not shown; however, the extent of the multimeric state is not known at the high concentration used for crystal formation). These results suggest that hpol may assemble to multimers in vivo, and the two mutants showed an increased tendency to multimeric assembly compared with wild-type hpol (Fig. 9), although extensive cellular studies would be needed for further conclusions about physiological relevance. The reason for this change in behavior is unknown, considering that the mutations are in the interior of the protein. Interestingly, multimeric forms of the other polymerases have been reported previously. For example, S. solfataricus Dpo4 can form a dimer on DNA in vitro (54). Small angle x-ray scattering and ultracentrifugation results reveal that mammalian DNA polymerase ␤ binds with DNA in ratios of both 2:1 and 1:1 in solution (55,56). Recently, an A-family DNA polymerase (polymerase ) has been reported to be able to form a dimer, which may be important for polymerase to participate in microhomology-mediated end joining (57). Considering those examples, it is possible that the multimeric form(s) of hpol plays a specific role in certain biological processes, although further investigation will be required.
hpol has been shown to have active site prealignment, dNTP-metal ion binding, and third metal ion binding steps during its catalytic cycle (34,38). In our study, the biochemical and structural results provide direct evidence that the conserved active site residues Arg-61 and Gln-38 synergistically contribute to enzyme efficiency and fidelity, further delineating the catalytic mechanism of the nucleotidyl transfer reaction catalyzed by hpol . . Comparison of wild-type hpol during the catalytic process with Q38A/R61A⅐8-oxoG⅐dCTP. A, the ternary structure of wild-type hpol during catalytic process (PDB code 4ECT (34)); B and C, active site views of superimposed structures of wild-type hpol complex and Q38A/R61A⅐8-oxoG⅐dCTP from the major groove and on top of the base pair.